Prediction Of Single And Mutiple Subcellular Location Of Apoptosis Proteins Based On Mrna And Fusion Information

With the rapid development of bioinformatics,appearing in a large number of unknown functional protein sequences,the study of the apoptosis protein attracts lots of attention from researchers.In a living organism,the abnormality of apoptosis proteins can trigger many kinds of diseases.The function of these proteins is closely related to their position in the cell.As a consequence,accurate prediction of the apoptosis protein subcellular localization can better understand its function.Here we constructed a new set of apoptosis proteins,including single-location dataset,multiple-location dataset and the mRNA-proteins dataset with mRNA information which used single-location apoptosis protein dataset as standard.For the prediction of single-location,it extracted physicochemical properties,stickiness and evolutionary information of apoptosis proteins through features screening,along with two mRNA information of apoptosis proteins:the three reading frame 3-mer mRNA sequence frequency information and mRNA secondary structure-sequence mode information.Combining all the information,using support vector machine algorithm,apoptosis proteins of four different subcellular localizations were predicted.The study found that the hybrid of mRNA and AAs information promoted prediction results.Under the Jackknife test,the overall prediction access rate achieved 82.18%while independent test datasets achieved 78.26%.Prediction results show that sequence and structure characteristics of the mRNA contribute to prediction of the subcellular localization of apoptosis proteins.For the prediction of multiple-location,taking the effect of the similarity of sequence on predicting results into consideration,the single-location dataset and multiple-location dataset with the similarity of sequence under 80%are chosen by a culling program CD-HIT.Several biological features,dipeptide composition,protein blocks composition,amino acid physicochemical properties and evolutionary information of apoptosis proteins,were effectively applied to predict the subcellular location of apoptosis protein by using Increment of Diversity algorithm.The overall localization accuracy of hybrid reached 79.57%.